Understanding Clinical Variability In Autism: Results From a California Twin Study

Background: Autism spectrum disorder (ASD) is characterized by a range of behavioral deficits in several domains.  One question is whether ASD represents a distinct clinical entity or simply the tail of a continuous distribution of behavioral traits.  A second question is the etiologic basis (genetic versus environmental) for the continuous gradient of severity observed in ASD.  Both questions can be addressed by a twin study.   

Objectives: To use objective measures of behavior based on direct assessment of MZ and DZ twins to determine whether clusters can be defined based on these measures that correlate with diagnosis, and the degree to which affected MZ and DZ twin pairs and discordant MZ and DZ pairs are correlated for these measures, and factor scores representing severity derived from them. 

Methods: 210 twin pairs (60 MZ; 150 DZ) with at least one twin carrying a clinical diagnosis of ASD were assessed using the Autism Diagnostic Observation Schedule (ADOS), Autism Diagnostic Interview-Revised (ADI-R), Stanford-Binet and/or Mullen Early Learning Scales, Vineland Adaptive Behaviors Scale (VABS), Social Responsiveness Scale (SRS), and Peabody Picture Vocabulary Test (PPVT). Statistical analysis included cluster analysis and factor analysis, as well as intraclass and interclass correlations and comparison of group means for quantitative variables (Full Scale IQ, Nonverbal IQ, Verbal IQ, VABS composite score, PPVT composite score, SRS T-score, and ADOS severity score.) 

Results: Cluster analysis of the quantitative variables first sorted twins into two distinct clusters that correlated very highly with ASD diagnosis.  Additional clusters were defined along a single continuously distributed severity gradient across all quantitative measures within the affected subjects. Factor analysis suggested that variability amongst the twins with ASD was largely attributable to a single factor based on IQ, VABS composite score, and PPVT composite score. For concordant ASD twin pairs, intraclass correlations of quantitative variables and the derived severity factor were highly significant and comparable in monozygotic (MZ) pairs (average r of 0.52) and dizygotic (DZ) pairs (average r of 0.479) . Correlations were not significantly higher in MZ as compared to DZ twins. Similarly, interclass correlations between affected and unaffected twins in discordant pairs for the quantitative variables were high and comparable for the MZ and DZ pairs. Comparisons of group means demonstrated that MZ twins appear to be more severely affected than DZ twins on all quantitative variables; this was true for both the affected and unaffected twins.

Conclusions: ASD appears to be a discrete clinical entity rather than the tail of a continuous behavior distribution.  However, there appears to be a single continuous severity gradient affecting all aspects of behavior characteristic of ASD among affected individuals.  Although twins correlate with one another with regards to severity, the correlations are relatively equal in MZ and DZ twins indicating that severity itself is not inherited. Rather, severity is likely due to environmental factors shared by twins. Additionally, MZ twins appear to be more severely affected than DZ twins suggesting that monozygosity may be an environmental risk factor for the development of ASD and its severity.